Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/1517#discussion_r70146784
--- Diff:
flink-scala/src/main/scala/org/apache/flink/api/scala/GroupedDataSet.scala ---
@@ -282,27 +283,57 @@ class GroupedDataSet[T: ClassTag](
}
/**
- * Creates a new [[DataSet]] by merging the elements of each group
(elements with the same key)
- * using an associative reduce function.
- */
+ * Creates a new [[DataSet]] by merging the elements of each group
(elements with the same key)
+ * using an associative reduce function.
+ */
def reduce(fun: (T, T) => T): DataSet[T] = {
+ reduce(getCallLocationName(), fun, CombineHint.OPTIMIZER_CHOOSES)
+ }
+
+ /**
+ * Special [[reduce]] operation for explicitly telling the system what
strategy to use for the
+ * combine phase.
+ * If null is given as the strategy, then the optimizer will pick the
strategy.
+ */
+ def reduce(fun: (T, T) => T, strategy: CombineHint): DataSet[T] = {
--- End diff --
I think it is OK to add overloaded `reduce()` methods to `DataSet` and
`GroupedDataSet`. These methods should be `PublicEvolving`. I would not add
overloaded methods to more specialized operations as in the first approach to
add the `CombineHint` to the Java API.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
